Paper Clips Run Amok

A simple thought experiment to shed light on the potential dangers of AI. Can we stop the earth being buried in paperclips?

We all know paperclips are thin metal wire bent into shapes to make it easy to hold together two or more sheets of paper. But are they dangerous? Could a machine that makes paperclips hurt people?

The effort to make computers as intelligent as humans has led to a thought experiment called the Paperclip Maximizer. Imagine a computer designed to do what humans do to make paperclips. It would buy metal to turn into paperclips, create paperclip designs, buy machines to make paperclips, hire people, and so on. The computer would be given one task: create paperclips. Now imagine the computer also has the ability to learn from mistakes and to figure out possible future scenarios, all to ensure it can make paperclips.

The computer would learn to do its job well. When the computer ran out of metal, for example, it might try to find other raw materials that could be turned into metal. When that no longer worked, the computer might look in our solar system, then beyond.

The computer would do its job but earth would be buried in mountains of paperclips.

Could people stop the computer before it destroyed earth? No. The computer would think of that possibility early on, as part of figuring out how to make sure it can produce paperclips. It would identify all obstacles, humans included, and plan how to prevent them.

Obviously, creating artificial intelligence to complete a task is not enough. Machines also must have values shared by humans. The paperclip making computer values creating paperclips even if it means using up every last bit of metal on earth. However, the result of using all the metal on earth destroys what humans value most, their need to live, to preserve human life.

The Paperclip Maximizer is a way to debate the limits of artificial intelligence. Creating a human-like intelligence in a computer might be the last thing humans do.

Learn More

Paperclip Maximizer

https://wiki.lesswrong.com/wiki/Paperclip_maximizer

AI and the Paperclip Problem

https://voxeu.org/article/ai-and-paperclip-problem

The Way the World Ends: Not a Bang But a Paperclip

https://www.wired.com/story/the-way-the-world-ends-not-with-a-bang-but-a-paperclip/

Author

  • Tim Slavin

    Tim is an award-winning writer and technologist who enjoys teaching tech to non-technical people. He has many years experience with web sites and applications in business, technical, and creative roles. He and his wife have two kids, now teenagers, who are mad about video games.

Also In The February 2019 Issue

The craft world and the tech world collide in this fun, hands-on activity.

For twenty three years, since 1996, cars have used computers to control different parts of the car.

Synchronizing games with players from all around the world is no easy task.

A simple thought experiment sheds light on the dangers of AI. Can we stop the earth being buried in paperclips?

‘Files’ may be easy concepts for humans, but not for computers. What’s going on inside your operating system?

Two ways to play Minecraft with an overarching storyline. Experience the game like never before!

Helping kids fall in love with coding through Minecraft mods and Raspberry Pis.

A better, smoother way to direct your theatrical masterpiece.

How hackers can steal the messages you send over the internet, and how to tell if your communications are secure.

It’s not as simple as it seems. Can you solve this classic programming problem?

Ever wondered what happens when you connect to a website? Time to dive into the secrets of networking!

Links from the bottom of all the February 2019 articles, collected in one place for you to print, share, or bookmark.

Interesting stories about science and technology for February 2019.

Interested but not ready to subscribe? Sign-up for our free monthly email newsletter with curated site content and a new issue email announcement that we send every two months.

No, thanks!